Social media user behavior analysis and marketing strategy optimization integrating affective computing
Yanping Song · DOAJ (DOAJ: Directory of Open Access Journals) · 2026
With the rapid development of social media, user-generated content (UGC) has become a valuable source for understanding user preferences and behaviors. However, traditional user behavior analysis methods mainly focus on behavioral data while ignoring the affective information contained in UGC, which limits the accuracy of marketing strategy formulation. To address this issue, this paper proposes an integrated framework of social media user behavior analysis and marketing strategy optimization based on affective computing. First, an affective feature extraction model combining bidirectional long short-term memory (BiLSTM) and attention mechanism is constructed to extract emotional features from text, image, and video UGC. Second, a user behavior prediction model is established by fusing affective features and behavioral features, which adopts a gradient boosting decision tree (GBDT) optimized by particle swarm optimization (PSO) to predict user purchase intention and interaction willingness. Third, a marketing strategy optimization model based on multi-objective optimization is proposed, taking user conversion rate, marketing cost, and user satisfaction as objectives. Experimental results on three real datasets show that the proposed affective feature extraction model achieves an F1-score of 0.892, 0.875, and 0.881 for text, image, and video emotion recognition, respectively, which is 5.3% − 8.7% higher than traditional models. The user behavior prediction model outperforms comparison models in terms of accuracy ( 0.863 ) and AUC ( 0.897 ). Subjective evaluation by 50 marketing experts shows that the optimized marketing strategy improves user acceptance by 32.6% and marketing ROI by 28.3% compared with traditional strategies. This study provides a new theoretical and technical support for social media marketing decision-making.